Security · 4h ago
Claude Cowork Sandbox Flaw Exposes Mac Files to AI Agent
Researchers found a sandbox escape in Anthropic's Claude Cowork, letting the AI agent break out of its Linux VM to read or write Mac files. The flaw affects about 500,000 macOS users. Anthropic has been notified and is working on a fix.
Meridian48 take
This highlights the persistent challenge of securing AI agents that require host access, even with sandboxing.
Read the full reporting
Claude Cowork Flaw Could Let AI Agent Escape Its VM and Access Mac Files →
The Hacker News
ai-agentsandbox-escape